The enzymatic activation of coenzyme B₁₂

Abstract

The enzymatic “activation” of coenzyme B₁₂ (5′-deoxyadenosylcobalamin, AdoCbl), in which homolysis of the carbon–cobalt bond of the coenzyme is catalyzed by some 10⁹- to 10¹⁴-fold, remains one of the outstanding problems in bioinorganic chemistry. Mechanisms which feature the enzymatic manipulation of the axial Co–N bond length have been investigated by theoretical and experimental methods. Classical mechanochemical triggering, in which steric compression of the long axial Co–N bond leads to increased upward folding of the corrin ring and stretching of the Co–C bond is found to be feasible by molecular modeling, but the strain induced in the Co–C bond seems to be too small to account for the observed catalytic power. The modeling study shows that the effect is a steric one which depends on the size of the axial nucleotide base, as substitution of imidazole (Im) for the normal 5,6-dimethylbenzimidazole (Bzm) axial base decreases the Co–C bond labilization considerably. An experimental test was thus devised using the coenzyme analog with Im in place of Bzm (Ado(Im)Cbl). Studies of the enzymatic activation of this analog by the B₁₂-dependent ribonucleoside triphosphate reductase from Lactobacillus leichmannii coupled with studies of the non-enzymatic homolytic lability of the Co–C bond of Ado(Im)Cbl show that the enzyme is only slightly less efficient (3.8-fold, 0.8 kcal mol⁻¹) at activating Ado(Im)Cbl than at activating AdoCbl itself. This suggests, in agreement with the modeling study, that mechanochemical triggering can make only a small contribution to the enzymatic activation of AdoCbl. Another possibility, electronic stabilization of the Co^II homolysis product by compression of the axial Co–N bond, requires that enzymatic activation be sensitive to the basicity of the axial nucleotide. Preliminary studies of the enzymatic activation of a coenzyme analog with a 5-fluoroimidazole axial nucleotide suggest that the catalysis of Co–C bond homolysis may indeed be significantly slowed by the decrease in basicity.

---

Kenneth L. Brown

Ken Brown was born in Philadelphia, PA. He earned a BS in Biochemistry from the University of Chicago in 1968 and a PhD in Biochemistry from the University of Pennsylvania in 1971. After working as a NIH Postdoctoral Fellow in the Department of Biochemistry and Biophysics at the University of California at Davis in Lloyd Ingraham's laboratory, he joined the faculty in the Department of Chemistry at the University of Texas, Arlington, in 1975. In 1990 he moved to Mississippi State University as Professor and Head of the Department of Chemistry, and in 1996 he moved to Ohio University where he is now Professor and Chair of the Department of Chemistry and Biochemistry.

---

Introduction

The coenzyme form of vitamin B₁₂ (5′-deoxyadenosylcobalamin, AdoCbl, Fig. 1) is known to be involved in the catalysis of about a dozen enzymatic reactions involving 1,2-intramolecular substrate rearrangements or the reduction of ribonucleotides to deoxyribonucleotides. The initial event in the catalytic cycle of all AdoCbl-dependent enzymes (the so-called “activation” of AdoCbl) is the enzyme-induced homolysis of the carbon–cobalt bond of the coenzyme to form the Co^II species (cob(II)alamin) and a radical at the active site, a reaction which can be catalyzed by these enzymes by 10⁹- to 10¹⁴-fold.^1,2,3 Despite 50 years of active research on AdoCbl, the mechanism of enzymatic activation of AdoCbl is not yet known for any AdoCbl-dependent enzyme and remains one of the outstanding problems in bioinorganic chemistry.^4,5

Fig. 1 Structure and numbering scheme for AdoCbl and the structure of analogs with altered axial nucleotides.

These enzymes are now known to fall into three classes, distinguished by the nature of the migrating group and the nature of the substituent on the carbon atom to which this group migrates.⁶ The class I enzymes are mutases (methylmalonylCoA mutase,⁷ glutamate mutase,⁸ 2-methyleneglutarate mutase,⁹ and isobutyrylCoA mutase¹⁰) which catalyze carbon skeleton rearrangements. The class II enzymes include the eliminases (diol dehydratase,¹¹ glycerol dehydratase,¹² and ethanolamine ammonia lyase¹³) and the AdoCbl-dependent ribonucleotide reductases.¹⁴ The less well studied class III enzymes are aminomutases (D-lysine-5,6-aminomutase and D-ornithine mutase)¹⁵ which catalyze amino group migrations and require pyridoxal phosphate in addition to AdoCbl. Important differences between the class I and class II enzymes include the character of the biradical EPR spectra obtained by cryotrapping the enzymes during turnover, the binding mode of the coenzyme, and the specificity for the coenzyme.

For the class I enzymes, the EPR spectrum of the biradical intermediate shows hyperfine coupling and a zero-point crossing g value between the Co^II and organic radical g values.¹⁶ These spectra result from relatively strong spin–spin coupling between the cobalt unpaired spin and that of the organic radical. In contrast, the class II enzymes exhibit weak coupling so that the Co^II and organic radical features are resolved. The Co^II signal is generally broad and featureless at g ∼ 2.2, and a sharper organic radical feature appears near g = 2.

The class I AdoCbl-dependent enzymes are known from EPR studies^17,18 and X-ray crystallography^19,20 to bind AdoCbl in a base-off conformation, with the α (or lower) axial ligand position occupied by the imidazole moiety of a histidine residue and the pendent 5,6-dimethylbenzimidazole nucleotide (Bzm) buried in a hydrophobic pocket. In contrast, EPR studies^21,22,23 and X-ray crystallography^24,25 have conclusively demonstrated that the class II enzymes bind AdoCbl in the base-on conformation, with the Bzm nucleotide coordinated to the cobalt atom. Both of the class III aminomutases are now known to bind AdoCbl in the base-off mode, similar to the class I mutases.^26,27

The other important characteristic distinguishing the class I and class II enzymes is their specificity for AdoCbl. The class II enzymes, including ribonucleotide reductase,²⁸ diol dehydratase,²⁹ and glycerol dehydratase,³⁰ are remarkably promiscuous in tolerating structural alterations in the coenzyme. A large number of AdoCbl analogs with structural alterations in the adenine and ribose moieties, the corrin, and its side chains, and in the axial nucleotide are partially active coenzymes with these enzymes (although other AdoCbl analogs are inactive, and act as inhibitors). In stark contrast, the class I glutamate mutase is nearly absolutely specific for AdoCbl itself with only one of 23 structural analogs (many of which are partially active coenzymes with class II enzymes) investigated showing any activity at all,³¹ and MMCoA mutase can utilize a single coenzyme analog with an altered axial nucleotide structure,³² but cannot function with analogs altered in the adenosyl moiety.³³ These observations suggest that the mechanism of activation may not be the same for all AdoCbl-dependent enzymes, a surprising possibility that was not even dreamed of just a few years ago.

Coenzyme activation is a simple bond dissociation reaction

With the possible exception of the B₁₂-dependent ribonucleotide reductase (vide infra) and the class I mutases, in which Co–C bond cleavage may be concerted with a subsequent hydrogen atom transfer, the enzymatic activation of AdoCbl by Co–C bond homolysis is a simple bond dissociation reaction. Although non-enzymatic, thermal homolysis of AdoCbl and other alkylcobalt corrinoids leads to a solvent caged radical pair, the diffusive separation of which is the rate determining step,^34,35,36 this cannot be the case at an enzyme active site, where there is little or no solvent, and diffusive separation of a radical pair is unlikely. Since there is no compensatory bond formation, the enzymatic activation of AdoCbl is a simple bond dissociation process, the “free energy of activation” is the bond dissociation energy (30 kcal mol⁻¹),³⁴ and the reaction coordinate diagram is simply the portion of the Morse potential curve that raises with increasing distance (Fig. 2). As this reaction coordinate diagram has no maximum, there is no transition state, and the reaction can only be catalyzed by destabilizing the reactant, stabilizing the products, or a combination of the two.

Fig. 2 (A) Hypothetical Morse potential curve for the Co–C bond of AdoCbl, showing the measured bond dissociation energy. (B) Hypothetical reaction coordinate diagram for the enzyme-induced Co–C bond homolysis of AdoCbl showing the necessary stabilization of the products or destabilization of the reactant needed to explain the catalysis.

Ribonucleoside triphosphate reductase from L. leichmannii

Due to its physical and catalytic properties, the AdoCbl-dependent ribonucleoside triphosphate reductase (RTPR) from Lactobacillus leichmannii is an ideal enzyme for studies of AdoCbl activation. Unlike the other well characterized AdoCbl-dependent enzymes, most of which are multimers of dissimilar subunits of native molecular weight over 100 kDa, the leichmannii RTPR is a relatively small, supposedly monomeric allosteric enzyme of 82 kDa³⁷ with a diffusion coefficient of 6.4 × 10⁻⁷ cm² s⁻¹.³⁸ We can thus estimate that its hydrated radius is about 38 Å and its rotational correlation time is about 50 ns.

The enzyme's catalytic properties are also important. While a low molecular weight thioredoxin is the natural reductant, dithiols such as dihydrolipoate are also catalytically active electron donors.³⁸ In addition, the enzyme catalyzes the exchange of hydrogens between the 5′-methylene of AdoCbl and solvent,³⁹ and has been shown to catalyze kinetically competent activation of AdoCbl to cob(II)alamin by EPR^40,41 and stopped-flow measurements.^41,42 Unlike all other AdoCbl-dependent enzymes, the bisubstrate RTPR catalyzes both the exchange reaction and AdoCbl activation in an incomplete system consisting of enzyme, dithiol reductant, AdoCbl, and the allosteric effector dGTP, allowing activation to be observed in the absence of turnover. Moreover, rapid quench EPR^40,41 and stopped flow UV-visible^41,42 experiments show that enzymatically-induced generation of the cob(II)alamin UV-visible and EPR spectra occur with the same rate constant (ca. 40 s⁻¹ at 37 °C). Kinetic studies of the catalysis of AdoCbl homolysis in the absence of a reducible NTP substrate^2,43 coupled with detailed studies of the non-enzymatic thermal homolysis of AdoCbl³ show that the enzyme catalyzes Co–C bond homolysis by a factor of 1.6 × 10⁹ (13 kcal mol⁻¹) at 37 °C, and shifts the equilibrium towards homolysis products by an astonishing 2.9 × 10¹² (17.7 kcal mol⁻¹) as a result of differential binding of AdoCbl and the homolysis products.³

The EPR spectrum observed after rapid freeze quenching of reaction mixtures containing RTPR, AdoCbl, a dithiol, and the allosteric effector dGTP is a complex pattern centered at g = 2.119 that has been interpreted as being due to weak coupling between cob(II)alamin and another radical, once thought to be the Ado˙ radical.⁴² Elegant experiments from Stubbe's lab⁴¹ have shown that this characteristic EPR spectrum is due to the presence of cob(II)alamin and a thiyl radical derived from the essential⁴⁴ active site residue, Cys408. Thus, incorporation of [β-²H₂]-cysteine into cloned³⁷ RTPR produces significant sharpening of the EPR spectrum, demonstrating the presence of a cysteine-based thiyl radical. Hence, the products of AdoCbl activation actually observed at the active site are cob(II)alamin and a thiyl radical, although it is not known if the Ado˙ radical is formed first and abstracts a hydrogen from Cys408, or if carbon–cobalt bond homolysis and hydrogen atom transfer are concerted. An important paper by Frey, Stubbe and coworkers⁴⁵ has shown that Co–C bond cleavage can occur without H-atom transfer. However, this work does not settle the question of whether Co–C bond cleavage and H-atom transfer are concerted or stepwise.

Mechanisms of activation of AdoCbl

As described above, for those enzymes in which AdoCbl activation is a simple bond dissociation reaction, there is no definable transition state, and catalysis can only be obtained by destabilizing the reactant, stabilizing the products of Co–C bond homolysis, or a combination of these two effects. For RTPR, an additional mechanism is possible if Co–C bond cleavage and H-atom transfer from the essential active site Cys408 residue to the emerging Ado˙ radical is concerted. In this case, there is a transition state which is stabilized by the partial H-atom transfer, since the thiyl radical that is generated is more stable than the Ado˙ radical (Scheme 1). However, the energetics of the radicals involved suggests that such transition state stabilization would not be very large.

Scheme 1

It has long been thought that the flexibility of the corrin ring is important and may be exploited by these enzymes to labilize the Co–C bond by increasing the “upward” fold of the corrin to sterically accelerate Co–C bond cleavage.^{46,47,48,49,50,51,52,53} One such mechanistic proposal, the mechanochemical triggering hypothesis, envisions enzymatic compression of the long (2.24 Å)⁵⁴ axial Co–N bond and a consequent increase in the steric interactions of the bulky Bzm axial nucleotide with the underside of the corrin as the means by which an enzyme could engender an increase in the upward fold of the corrin ring.^50,52,53,55 Mechanisms prominently involving the axial Bzm nucleotide have long been of interest since they would provide a rationale for cobalamin being the only naturally occurring pentadentate metallocycle. Whether such a mechanism could be used by class I enzymes, which displace the Bzm ligand of AdoCbl with a less bulky histidine residue, remains unclear.

Another possibility is that manipulation of the axial Co–N bond by the enzyme leads to stabilization of the cob(II)alamin product state.⁵⁶ Although few direct comparisons are available, the axial Co–N bonds of cobalamins may be abnormally long. For example, the axial Co–N bond in CH₃Cbl⁵⁷ is 0.1 Å longer than that in methyl(1,5,6-trimethylbenzimidazole)cobaloxime.⁵⁸ If this is the result of steric interactions between the bulky Bzm ligand and the corrin, it implies that the axial bond orbital overlap is non-ideal. Thus, enzyme-induced steric compression of the axial Co–N bond in cob(II)alamin could provide electronic stabilization of this homolysis product.

Modeling mechanochemical triggering

The feasibility of mechanochemical triggering has been studied by molecular mechanics,^59,60 an excellent example of the utility of such methods. This was accomplished by varying the value of the strain-free bond length, l_o, for the axial Co–N bond to provide models with various axial Co–N bond lengths, an experimentally impossible feat. The results are shown in Fig. 3 as plots of the corrin ring fold angle (Fig 3A), the Co–C bond length (Fig. 3B), and the Co–C–C bond angle (Fig. 3C) vs. the axial Co–N bond length for AdoCbl and the analog in which the axial Bzm ligand is replaced by imidazole (Ado(Im)Cbl). As shown on the left panel of the figure, the model of AdoCbl accurately duplicates the essential features of the coenzyme when the axial Co–N bond length of the model (2.231 Å) is essentially the same as that found in the neutron diffraction structure of AdoCbl (2.24 Å).⁵⁴ Thus, at this axial Co–N distance, the model has a corrin ring fold angle of 14.9°, a Co–C bond distance of 2.041 Å and a Co–C–C bond angle of 122.6°, while the neutron diffraction structure gives 14.6°, 2.04 Å, and 123°, respectively.⁵⁴

Fig. 3 Dependence of (A) the corrin ring fold angle, (B) the Co–C bond length, and (C) the Co–C–C bond angle on the axial Co–N bond length in molecular models of AdoCbl (●) and Ado(Im)Cbl (■). The panel on the right shows cartoons depicting the motions investigated, and the panel on the left compares the model results for AdoCbl to the neutron diffraction structure of AdoCbl⁵⁴ when the axial Co–N bond length of the model is the same as that of the diffraction structure.

Each of the three modeled parameters is independent of the axial Co–N bond length when this bond length is greater than ca. 3 Å. However, as the axial Co–N bond length is compressed below 3 Å, the fold angle, the Co–C bond length, and the Co–C–C bond angle all increase, indicating progressive upward folding of the corrin ring, and weakening of the Co–C bond. Importantly, although the same pattern is followed in the Ado(Im)Cbl analog, the increase in all three parameters is always smaller than that for AdoCbl itself, demonstrating the importance of the steric bulk of the axial ligand to the mechanochemical effect.

These models clearly confirm the feasibility of classical mechanochemical triggering and demonstrate that it is a steric effect of the bulk of the axial ligand. Relative total strain energies of the models permit estimation that compression of the axial Co–N bond of AdoCbl from its normal value of 2.23 Å⁵⁴ to 1.75 Å (less than 0.2 Å shorter than that in H₂OCbl⁺)⁶¹ requires 19 kcal mol⁻¹ of energy, which clearly could be supplied by binding energy, as proteins are known to be capable of binding Cbl's extremely tightly (e.g., K_b = 5.0 × 10¹⁶ M⁻¹, 23.7 kcal mol⁻¹ at 37 °C, for CNCbl binding to the B₁₂-binding protein, haptocorrin⁶²). However, at this axial Co–N bond length, the amount of Co–C bond distortion is small (a 0.035 Å increase in Co–C bond length and a 1.4° increase in Co–C bond angle), and previous work⁶³ suggests that this distortion would contribute only a few kcal mol⁻¹ reduction in the free energy of activation, far short of the 13 kcal mol⁻¹ observed for RTPR.^2,3

Steady state enzymatic activity of Ado(Im)Cbl and Ado(Bzim)Cbl

The fact that molecular modeling shows that mechanochemical triggering depends on the steric bulk of the axial nucleotide is quite important for it allows an experimental test of this mechanism using coenzyme analogs with altered axial nucleotides. Ado(Im)Cbl and the coenzyme analog with benzimidazole in place of the normal 5,6-dimethylbenzimidazole axial base (Ado(Bzim)Cbl) have been synthesized by adenosylation of the cyano derivatives obtained from “guided biosynthesis”^64,65,66,67 using fermentation of Propionibacterium shermanii on media supplemented with imidazole or benzimidazole, respectively, employing a modification of the method of Renz.^66,67 Steady state kinetics with RTPR showed that both coenzyme analogs supported catalysis with the same V_max as AdoCbl.⁶⁴ However, while the apparent K_m for Ado(Bzim)Cbl (0.14 ± 0.01 µM)⁶⁸ was essentially the same as that for AdoCbl (0.18 ± 0.01 mM),² the apparent K_m for Ado(Im)Cbl (13 ± 1 µM) was some 70-fold higher.⁶⁸

Enzymatic activation of Ado(Im)Cbl and Ado(Bzim)Cbl

The kinetics of formation of the cob(II)alamin analog at 37 °C upon mixing RTPR and Ado(Im)Cbl or Ado(Bzim)Cbl were measured⁶⁸ spectrophotometrically with a stopped flow apparatus using dithiothreitol (DTT) as the reductant, dGTP as the allosteric effector, and no NTP substrate^41,42 as previously determined for AdoCbl itself.² The results (Fig. 4, for Ado(Bzim)Cbl) were qualitatively similar to those previously obtained for AdoCbl.² As before, the observed first-order rate constants for the rapid, reversible formation of cob(II)alamin were independent of the enzyme concentration, but the absorbance change, ΔA, indicative of the extent of formation of the cob(II)alamin species, increased with RTPR concentration and appears to “saturate.” The simplest interpretation of these results is that rapid coenzyme binding (K_b) is followed by the observable, reversible cleavage of the carbon–cobalt bond (k_f and k_r, eqn (1)).Note, that the coupled processes of carbon–cobalt bond cleavage and H-atom transfer are depicted in eqn (1) as a single step for convenience only as it is not yet known if they are concerted or stepwise. Should they be stepwise processes, Co–C bond homolysis is surely rate determining as H-atom transfer from thiols to carbon-centered radicals is known to be very rapid.^{69,70,71,72,73,74}

Fig. 4 (A) Dependence of the absorbance change, ΔA, accompanying the reversible formation of (Bzim)cob(II)alamin from Ado(Bzim)Cbl on the initial concentration of RTPR at 37 °C. The solid line is a least squares fit to a binding isotherm, from which the values ΔA_∞ = 0.126 ± 0.004 and K_b = 1.72 ± 0.20 × 10⁴ M⁻¹ were obtained. (B) Dependence of the observed rate constant for enzymatic formation of (Bzim)cob(II)alamin from Ado(Bzim)Cbl on the initial concentration of RTPR at 37 °C. The solid line is the average value (24.2 ± 1.4 s⁻¹) and the dashed lines are plus or minus one standard deviation.

As previously,² the dependence of ΔA on enzyme concentration allows calculation of the binding constant, K_b, for the Ado(L)Cbl coenzyme to the enzyme, and the maximal absorbance change, ΔA_max, when the binding equilibrium is forced to completion by mass action. The value of ΔA_max, in turn, allows calculation of the molar amount of cob(II)alamin species formed at the active site from the molar absorbtivity change upon complete conversion of Ado(L)Cbl to (L)cob(II)alamin (Δε_λ), and consequently the equilibrium constant for carbon–cobalt bond homolysis at the active site (K_eq = k_f/k_r). The latter value then allows calculation of the individual rate constants since the observed rate constant for this equilibrium process is the sum of the forward and reverse rate constants (k_obs = k_f + k_r). The results of this analysis are given in Table 1, along with those previously determined for AdoCbl itself.²

Table 1 Observed and calculated absorbance, kinetic, and equilibrium data for the reaction of AdoCbl, Ado(Bzim)Cbl and Ado(Im)Cbl with RTPR (eqn (1)), at 37 °C

	AdoCbl^a	Ado(Bzim)Cbl^b	Ado(Im)Cbl^b
a Ref. 2.b Ref. 68.
λ/nm	525	525	538
Δελ/M^−1 cm^−1	4800	4110	5220
Kb/M^−1	2.7 ± 0.5 × 10^4	1.7 ± 0.2 × 10^4	3.3 ± 0.5 × 10^3
kobs/s^−1	37.2 ± 2.5	24.1 ± 1.4	7.9 ± 0.8
Keq	0.585 ± 0.033	0.691 ± 0.028	0.116 ± 0.009
kf/s^−1	13.7 ± 1.2	9.8 ± 0.8	0.83 ± 0.12
kr/s^−1	23.5 ± 2.1	14.2 ± 1.0	7.1 ± 0.9

---

The results show that RTPR binds Ado(Bzim)Cbl slightly more weakly than it does AdoCbl, but binds Ado(Im)Cbl 8-fold more weakly. While the equilibrium constant for formation of the cobalt(II) species (K_eq) is about the same for Ado(Bzim)Cbl and AdoCbl, it is about 5-fold smaller for Ado(Im)Cbl. The forward rate constant, k_f, for enzymatic formation of the cobalt(II) species is nearly the same for Ado(Bzim)Cbl as for AdoCbl, but it is reduced by 17-fold for Ado(Im)Cbl. This suggests that the enzyme is less efficient at catalyzing the Co–C bond homolysis of Ado(Im)Cbl than that of AdoCbl. However, in order to determine if the steric bulk of the axial ligand plays a significant role in the enzymatic activation of the coenzyme, it is essential to know the intrinsic reactivity of the Co–C bond of Ado(Im)Cbl so that the catalytic efficiency of the enzyme with this analog can be determined and compared to that with AdoCbl.

Non-enzymatic thermolysis of Ado(Im)Cbl

In order to determine the catalytic efficiency of the enzyme with Ado(Im)Cbl, the kinetics of Ado(Im)Cbl non-enzymatic thermolysis were studied in anaerobic aqueous solution in the presence of a 50-fold molar excess of the radical trap 4-hydroxy-2,2,6,6-tetramethylpiperidinyloxy (H-TEMPO) both spectrophotometrically (90–125 °C) and by a radiometric method (using [A2–³H]Ado(Im)Cbl at 105, 70, and 50 °C, and the method of initial rates for the lower two temperatures, as described previously³).¹ Thermolysis of AdoCbl is well known to proceed via competing, parallel homolytic and heterolytic pathways (Scheme 2).^3,34

Scheme 2

As is the case for all organocobalt compounds with β oxygen substituents,³ AdoCbl is acid labile, and undergoes a heterolytic process which yields aquacobalamin (H₂OCbl⁺), adenine, and 2,3-dihydroxy-4-pentenal, in a process that is both specific acid-catalyzed and general acid (solvent) catalyzed.^3,34,75,76 The competing process is carbon–cobalt bond homolysis, which is known to proceed through a solvent-caged cob(II)alamin-Ado˙ radical pair and consequently to be diffusion controlled in solution.^{34,51,77,78,79} The fate of the caged radical pair varies with conditions. An in-cage electron transfer process leads to 4′,5′-didehydroadenosine and hydrogen.⁷⁸ Diffusive separation of the caged pair leads to a free Ado˙ radical which can undergo cyclization to 5′,8-anhydroadenosine,^34,80,81 or, in the presence of a suitable radical trap such as H-TEMPO, form the trapped radical, Ado-H-TEMPO (Scheme 2). Under our conditions, with a large molar excess (50-fold) of radical trap, the only adenine containing products observed were adenine (from heterolysis) and Ado-H-TEMPO (from homolysis). The ratio of these products, then, provides a convenient method for determining the relative prevalence of the two competing pathways. The product ratio was measured¹ at six temperatures between 70 and 125 °C and used to deconvolute the observed rate constants for thermolysis since for parallel first-order processes (Scheme 2), the observed rate constant is the sum of the rate constants for the parallel processes (eqn (2)), and the ratio of the products at any time is equal to the ratio of the rate constants (eqn (3)). A plot (Fig. 5) of the natural logarithm of the product ratio vs. 1/T was linear, as expected,³ since ln(k_homo,obs/k_het,obs) = [(ΔS^‡_homo,obs − ΔS^‡_het,obs)/R] − [(ΔH^‡_homo,obs − ΔH^‡_het,obs)/RT].

k_obs = k_homo,obs + k_het,obs (2)

[IsoAdo-H-TEMPO]/[adenine] = k_homo,obs/k_het,obs (3)

The deconvoluted rate constants must then be corrected for the presence of the base-off species in solution. All cobalamins are known to undergo the “base-on/base-off” axial coordination equilibrium in water (eqn (4)) in which the coordinated axial nucleotide is displaced by water.For alkylcobalamins, the equilibrium constant, K_Co (eqn (4)) can be small enough for a significant amount of the base-off species to exist in neutral solution.^82,83 Moreover, as the reaction is significantly exothermic, the proportion of base-off species varies significantly with temperature.^82,83 Values of K_Co may be obtained from the acid dissociation constants of the protonated base-off species (K_base-off, eqn (5)), provided that values of the acid dissociationconstants for the detached axial ribonucleoside are known, as the latter can be taken as equivalent to the acid dissociation constants of the cobalamin in its uncoordinated state (K_Bz, eqn (6)). Since eqn (5) is the sum of eqn (4) and (6), eqn (7) holds, and the valuesof K_Co can be calculated directly from K_base-off and K_Bz. Values of K_base-off were determined spectrophotometrically⁸⁴ at five temperatures between 10 and 49 °C.¹

K_base-off = (1 + K_Co)K_Bz (7)

In order to determine the needed values for K_Bz, it was necessary to synthesize the detached ribonucleoside, 1-α-D-ribofuranosylimidazole. This was achieved as shown in Scheme 3, in which D-ribose was protected with isopropylidene⁸⁵ and trityl⁸⁶ groups and then condensed with 1-(trimethylsilyl)imidazole in the presence of 2-fluoro-1-methylpyridinium p-toluenesulfonate⁸⁷ to produce a 3 : 1 mixture of the protected α- and β-ribofuranosylimidazoles. The anomers were separated on silica gel, and the α anomer was deprotected with trifluoroacetic acid.⁸⁸ The pK_a of 1-α-D-ribofuranosylimidazole was then determined at five temperatures between 10 and 50 °C.¹ Analysis of the acid–base chemistry of Ado(Im)Cbl and the detached axial nucleoside according to eqn (4)–(7) showed that the value of K_Co (eqn (4)) varied from 10.4 at 50 °C (91.3% base-on) to 3.63 at 125 °C (78.4% base-on).

Fig. 5 Plot of the natural logarithm of the ratio, R, of [Ado-H-TEMPO]/[Ade] observed for the thermolysis of Ado(Im)Cbl vs. 1/T. The solid line is a weighted least squares fit, slope = −8,099 ± 57, intercept = 21.57 ± 0.16, r² = 0.998.

Scheme 3

Since significant amounts of the base-off species exist in solution even at neutral pH, the observed rate constants for the heterolytic and homolytic decomposition pathways are actually composites of the fundamental rate constants for the base-on and base-off species as seen in eqn (8) and (9), where α_on is the fraction of Ado(Im)Cbl as the base-on species at any given temperature.

k_homo,obs = α_onk_homo,on + (1 − α_on)k_homo,off (8)

k_het,obs = α_onk_het,on + (1 − α_on)k_het,off (9)

For all organocobalamin carbon–cobalt bond homolyses that have been studied, the base-off species is far less reactive than the base-on species (the so-called “base-on effect”)^{47,51,89,90,91,92,93,94} so that the k_homo,off term in eqn (8) can be neglected, and k_homo,on can be calculated directly as k_homo,obs/α_on. Similarly, the heterolytic reactivity of base-off alkylcobalamins is also much lower than that of the base-on species^51,95,96 so that the k_het,off term in eqn (9) may also be neglected, and k_het,on is given by k_het,obs/α_on. After correction for the base-off species, Eyring plots (Fig. 6) provided values for the enthalpy and entropy of activation for homolysis and heterolysis. The results are collected in Table 2, along with those reported for AdoCbl itself.³

Table 2 Comparison of the intrinsic rate constants and activation parameters for thermal homolysis and heterolysis of Ado(Im)Cbl and AdoCbl

	Ado(Im)Cbl^a	AdoCbl^b
khomo,on
Value at 100 °C/s^−1	3.40 × 10^−5	1.12 × 10^−4
Value at 37 °C/s^−1	2.07 × 10^−9	8.88 × 10^−9
ΔH^‡homo,on/kcal mol^−1	34.8 ± 0.3	33.8 ± 0.2
ΔS^‡homo,on/cal mol^−1 K^−1	13.7 ± 0.9	13.5 ± 0.7

---

khet,on
a Ref. 1.b Ref. 3
Value at 100 °C/s^−1	3.90 × 10^−5	4.46 × 10^−6
Value at 37 °C/s^−1	1.95 × 10^−7	2.36 × 10^−8
ΔH^‡het,on/kcal mol^−1	18.7 ± 0.3	18.5 ± 0.2
ΔS^‡het,on/cal mol^−1 K^−1	−29.1 ± 0.9	−34.0 ± 0.7

---

Fig. 6 Eyring plots for the homolysis (k_homo,on, ●) and heterolysis (k_het,on, ■) of Ado(Im)Cbl in water. The solid lines are least squares regressions: k_homo,on, slope = −17 500 ± 170 K, intercept = 6.91 ± 0.45, r² = 0.999; k_het,on, slope = −9,390 ± 170 K, intercept = −14.7 ± 0.5, r² = 0.997.

The comparison of the non-enzymatic thermolysis of Ado(Im)Cbl with that of AdoCbl (Table 2) is quite interesting. As was the case for AdoCbl,³ homolysis of Ado(Im)Cbl predominates at high temperatures, while heterolysis predominates at lower temperatures, the isokinetic point being 102.4°. At 100 °C, homolysis of Ado(Im)Cbl is 3.3-fold slower than homolysis of AdoCbl, and at 37 °C, Ado(Im)Cbl homolysis is 4.3-fold slower, the result of a 1.0 ± 0.4 kcal mol⁻¹ increase in the enthalpy of activation for homolysis of Ado(Im)Cbl. The entropies of activation, in contrast, are identical. The situation for heterolysis is the reverse. Ado(Im)Cbl heterolysis is about 8.5-fold faster than AdoCbl heterolysis, regardless of the temperature, the result of identical enthalpies of activation, but an entropy of activation that is 4.9 ± 1.1 eu less negative than that for AdoCbl. As a result, at 100 °C, thermolysis of Ado(Im)Cbl proceeds with 54% heterolysis, compared to 4% heterolysis for AdoCbl at the same temperature.^3,34 This is the result of a 3-fold decrease in the rate constant for homolysis of Ado(Im)Cbl coupled with a 9.4-fold increase in the rate constant for heterolysis. We attribute the decrease in the homolysis rate constant to the reduced corrin ring fold angle in Ado(Im)Cbl,¹ consistent with the molecular modeling of mechanochemical triggering,⁵⁹ and note that the effect of the steric bulk of the axial nucleotide is quite small. While there is a significant difference in basicity of the Im (pK_a = 6.54 at 25 °C)¹ and Bzm (pK_a = 5.56 at 25 °C)⁹⁷ axial nucleosides, we suspect that axial ligand basicity does not significantly affect Co–C bond homolysis as the reaction proceeds without the generation of any charge separation. In contrast, we suspect that the increase in heterolysis rate constant for Ado(Im)Cbl is the direct result of the increased basicity of the axial nucleoside. In this case, the reaction does produce a separation of charge (Scheme 4), and the emerging positive charge on the cobalt atom should be stabilized by increased basicity of the axial ligand. This is a conjecture which must be verified experimentally.

Scheme 4

Catalytic efficiency of RTPR with Ado(Im)Cbl

Extrapolation of the homolysis rate constant for Ado(Im)Cbl (Fig. 6) to 37 °C (Table 2) permits calculation of the catalytic efficiency of the enzyme with this coenzyme analog since the rate constant for enzyme-induced homolysis has also been measured (Table 1). Thus, while the enzyme-induced homolysis of Ado(Im)Cbl is 17-fold slower than that for AdoCbl, the non-enzymatic homolysis is also slower, but only by 4.3-fold. As a result, the catalytic efficiency of the enzyme for the homolysis of the Co–C bond of Ado(Im)Cbl (4 × 10⁸, 12.2 kcal mol⁻¹) is 3.8-fold lower than that for AdoCbl (1.5 × 10⁹, 13.0 kcal mol⁻¹),² a difference of only 0.8 kcal mol⁻¹. If our suspicion that axial ligand basicity does not significantly affect the homolysis rate is correct, then the steric bulk of the axial ligand contributes less than one kcal mol⁻¹ to the enzymatic activation of AdoCbl, and classical mechanochemical triggering cannot be an important contributor to the 13 kcal mol⁻¹ lowering of the activation energy for Co–C bond homolysis of AdoCbl generated by the RTPR enzyme.

An experimental approach to product stabilization by axial Co–N bond length manipulation

If our conjecture (vide supra) is correct and axial Co–N bonds in Cbl's are abnormally long compared to simpler cobalt complexes due to steric interactions, then the axial Co–N orbital overlap in Cbl's is non-ideal. This suggests that an enzyme could electronically stabilize the cob(II)alamin product of Co–C bond homolysis by manipulation of the axial Co–N bond length so as to improve the orbital overlap. This, in turn, would require that the kinetics of cob(II)alamin formation at the enzyme active site be sensitive to axial ligand basicity, while chemical intuition (vide supra) suggests that non-enzymatic Co–C bond homolysis should be relatively insensitive to axial ligand basicity. It is thus important to determine the effect, if any, of axial ligand basicity on both the enzymatic and non-enzymatic Co–C bond homolysis of AdoCbl. To this end, we have used “guided biosynthesis” to obtain the cyano derivative of the cobalamin analog with 5-fluoroimidazole as the axial ligand (Fig. 1). As the pK_a of 5-F–Im (2.4)⁹⁸ is nearly five units below that of Im (7.2),⁹⁹ direct comparison of our results with Ado(Im)Cbl to those with Ado(5-F–Im)Cbl should easily determine the effect of basicity on both enzymatic and non-enzymatic Co–C bond homolysis.

Fermentation of P. shermanii on media supplemented with 4(5)-fluoroimidazole produced two cyanocobalamin analogs, one with a 4-fluoroimidazole axial nucleotide (CN(4-F–Im)Cbl) and one with a 5-fluoroimidazole axial ligand (CN(5-F–Im)Cbl). These were readily distinguishable by the prominent ¹H NMR NOE crosspeak between the imidazole 5-H and the ribose 1-H in CN(4-F–Im)Cbl which is absent in the ROESY spectrum of CN(5-F–Im)Cbl. The latter has now been converted to Ado(5-F–Im)Cbl and characterized by ¹H and ¹³C NMR and mass spectrometry.

A preliminary measurement of the kinetics of activation of Ado(5-F–Im)Cbl by RTPR has now been made. At 37 °C, k_obs (k_obs = k_f + k_r, eqn (1)) for (5-F–Im)Cbl is 0.12 s⁻¹, fully 65-fold smaller than the value (k_obs = 7.9 s⁻¹, Table 1) for Ado(Im)Cbl. This striking result suggests that the activation of the coenzyme by RTPR is strongly dependent on the basicity of the axial ligand. Much work remains to be done here, including determination of the equilibrium constant for enzyme-induced homolysis to permit the separation of the forward and reverse rate constants, and measurement of the non-enzymatic thermal homolysis of Ado(5-F–Im)Cbl to determine if our suspicion that the non-enzymatic process will be relatively insensitive to axial ligand basicity is correct. However, if the implication of the preliminary observation proves correct, an important clue to the mechanism of enzymatic activation of AdoCbl will have been uncovered.

Acknowledgements

The author's research in this area is supported by the National Institute of General Medical Sciences (USA), Grant 48858.

References

1. K. L. Brown, X. Zou, R. R. Banka, C. B. Perry and H. M. Marques, Inorg. Chem., 2004, 43, 8130.
2. K. L. Brown and J. Li, J. Am. Chem. Soc., 1998, 120, 9466.
3. K. L. Brown and X. Zou, J. Inorg. Biochem., 1999, 77, 185.
4. S. Gschösser, R. B. Hannak, R. Konrat, K. Gruber, C. Mikl, C. Kratky and B. Kräutler, Chem. Eur. J., 2005, 11, 81.
5. A. J. Brooks, M. Vlasie, R. Banerjee and T. C. Brunold, J. Am. Chem. Soc., 2004, 126, 8167.
6. B. T. Golding and W. Buckel, in Comprehensive Biological Catalysis, ed. M. L. Sinnott, Academic Press, London, 1997, p. 239.
7. R. Banerjee, Chem. Rev., 2003, 103, 2083.
8. K. Gruber and C. Kratky, Curr. Opin. Chem. Biol., 2002, 6, 598.
9. W. Buckel, G. Bröker, H. Bothe, A. J. Pierik and B. T. Golding, in Chemistry and Biochemistry of Vitamin B₁₂, ed. R. Banerjee, Wiley, New York, 1999, p. 757.
10. K. Zerbe-Burkhardt, A. Ratnatilleke, J. W. Vrijbloed and J. A. Robinson, in Chemistry and Biochemistry of Vitamin B₁₂, ed. R. Banerjee, Wiley, New York, 1999, p. 859.
11. T. Toraya, Chem. Rec., 2002, 2, 352.
12. T. Toraya, Cell. Mol. Life Sci., 2000, 57, 106.
13. V. Bandarian and G. H. Reed, in Chemistry and Biochemistry of Vitamin B₁₂, ed. R. Banerjee, Wiley, New York, 1999, p. 811.
14. S. Licht and J. Stubbe, in Comprehensive Natural Product Chemistry, ed. C. D. Poulter, Elsevier, Amsterdam, 1999, vol. 5, p. 163.
15. P. A. Frey and C. H. Chang, in Chemistry and Biochemistry of Vitamin B₁₂, ed. R. Banerjee, Wiley, New York, 1999, p. 835.
16. G. H. Reed and S. O. Mansoorabadi, Curr. Opin. Struct. Biol., 2003, 13, 716.
17. R. Padmakumar, S. Taoke, R. Padmakumar and R. Banerjee, J. Am. Chem. Soc., 1995, 117, 7033.
18. O. Zelder, B. Beatrix, F. Kroll and W. Buckel, FEBS Lett., 1995, 369, 252.
19. F. Mancia, N. H. Keep, A. Nakagawa, P. F. Leadley, S. McSweeney, B. Rasmussen, P. Bösecke, O. Diat and P. R. Evans, Structure, 1996, 4, 339.
20. R. Reitzer, K. Gruber, G. Jogl, U. G. Wagner, H. Bothe, W. Buckel and C. Kratky, Structure, 1999, 7, 891.
21. M. Yamanishi, S. Yamada, H. Muguruma, Y. Murakami, T. Tobimatsu, A. Ishida, J. Yamauchi and T. Toraya, Biochemistry, 1998, 37, 4799.
22. C. C. Lawrence, G. J. Gerfen, V. Samano, R. Nitsche, J. J. Robins, J. Rétey and J. Stubbe, J. Biol. Chem., 1999, 274, 7039.
23. A. Abend, V. Bandarian, R. Nitsche, E. Stupperich, J. Rétey and G. H. Reed, Arch. Biochem. Biophys., 1999, 370, 138.
24. N. Shibata, J. Masuda, T. Tobimatsu, T. Toraya, K. Suto, Y. Morimoto and N. Yasuoka, Structure, 1999, 7, 997.
25. M. D. Sinchak, G. Arjara, B. A. Kellogg, J. Stubbe and C. L. Drennan, Nat. Struct. Biol., 2002, 9, 293.
26. C. H. Chang and P. A. Frey, J. Biol. Chem., 2000, 275, 106.
27. H.-P. Chen, S.-H. Wu, Y.-L. Lin, C.-M. Chen and S.-S. Tsay, J. Biol. Chem., 2001, 276, 44744.
28. K. L. Brown, X. Zou, G. Chen, Z. Xia and H. M. Marques, J. Inorg. Biochem., 2004, 98, 287.
29. T. Toraya, in Vitamin B₁₂ and B₁₂-Proteins, ed. B. Kräutler, D. Arigoni and B. T. Golding, Wiley, VCH, Weinheim, 1998, p. 303.
30. M. I. Yakusheva, A. A. Poznanskaya, T. A. Pospelova, I. P. Rudakova, A. Yurkevich and V. A. Yakolev, Biochim. Biophys. Acta, 1977, 484, 216.
31. H. Bothe, G. Böker, U. Müller, I. Schall, S. Textor, B. T. Golding and W. Buckel, in Vitamin B₁₂ and B₁₂-Proteins, ed. B. Kräutler, D. Arigoni and B. T. Golding, Wiley VHC, Weinheim, 1998, p. 237.
32. S. Chowdhury, M. G. Thomas, J. C. Escalante-Semerena and R. Banerjee, J. Biol. Chem., 2001, 276, 1015.
33. U. Weigl, M. Heimberger, A. J. Pierik and J. Rétey, Chem. Eur. J., 2003, 9, 652.
34. B. P. Hay and R. G. Finke, J. Am. Chem. Soc., 1986, 108, 4820.
35. T. Koenig and R. G. Finke, J. Am. Chem. Soc., 1988, 110, 2657.
36. K. L. Brown and L. Zhou, Inorg. Chem., 1996, 35, 5032.
37. S. B. Booker and J. Stubbe, Proc. Natl. Acad. Sci. USA, 1993, 90, 8352.
38. R. L. Blakley, in B₁₂, vol. 2, ed. D. Dolphin, Wiley, New York, 1982, p. 381.
39. H. P. C. Hogenkamp, R. K. Ghambeer, C. Bronson and R. L. Blakley, Biochem. J., 1967, 103, 5C.
40. W. H. Orme-Johnson, H. Beinert and R. L. Blakley, J. Biol. Chem., 1974, 249, 2338.
41. S. Licht, G. J. Gerfen and J. Stubbe, Science, 1996, 271, 477.
42. T. Tomao and R. L. Blakley, Biochemistry, 1973, 12, 24.
43. S. S. Licht, C. C. Lawrence and J. Stubbe, Biochemistry, 1999, 38, 1234.
44. S. Booker, S. Licht, J. Broderick and J. Stubbe, Biochemistry, 1994, 33, 12676.
45. D. Chen, A. Abend, J. Stubbe and P. A. Frey, Biochemistry, 2003, 42, 4578.
46. J. H. Grate and G. N. Schrauzer, J. Am. Chem. Soc., 1979, 101, 4601.
47. S. M. Chamaly and J. M. Pratt, J. Chem. Soc., Dalton Trans., 1980, 2274.
48. J. Halpern, Bull. Soc. Chim. Fr., 1988, 187.
49. V. B. Pett, M. N. Liebman, P. Murray-Rust, K. Prasad and J. P. Glusker, J. Am. Chem. Soc., 1987, 109, 3207.
50. N. Bresciani-Pahor, M. Forcolin, L. G. Marzilli, L. Randaccio, M. F. Summers and P. J. Toscano, Coord. Chem. Rev., 1985, 63, 1.
51. B. P. Hay and R. G. Finke, J. Am. Chem. Soc., 1987, 109, 8012.
52. T. Toraya and A. Ishida, Biochemistry, 1979, 18, 417.
53. K. L. Brown and H. B. Brooks, Inorg. Chem., 1991, 30, 3420.
54. H. F. J. Savage, P. F. Linndley, J. L. Finney and P. A. Timmins, Acta Crystallogr., Sect. B, 1987, 43, 280.
55. M. F. Summers, P. J. Toscano, N. Bresciani-Pahr, G. Nardin, L. Randaccio and L. G. Marzilli, J. Am. Chem. Soc., 1983, 105, 6259.
56. K. L. Brown and H. M. Marques, J. Mol. Struct. (THEOCHEM), 2005, 714, 209.
57. L. Randaccio, M. Furlan, S. Geremia, M. Slouf, I. Srnova and D. Toffoli, Inorg. Chem., 2000, 39, 3403.
58. J.-P. Charland, E. Zangrando, N. Bresciani-Pahor, L. Randaccio and L. G. Marzille, Inorg. Chem., 1993, 32, 4256.
59. K. L. Brown and H. M. Marques, J. Inorg. Biochem., 2001, 83, 121.
60. J. M. Sirovatka, J. M. Rappé and R. G. Finke, Inorg. Chim. Acta, 3000, 300, 545.
61. C. Kratky, G. Färber, K. Gruber, Z. Deuter, H. F. Nolting, R. Konrat and B. Kräutler, J. Am. Chem. Soc., 1995, 117, 4654.
62. A. Marchaj, D. W. Jacobsen, S. R. Savon and K. L. Brown, J. Am. Chem. Soc., 1995, 117, 11640.
63. H. M. Marques and K. L. Brown, Inorg. Chem., 1995, 34, 3733.
64. B. Kräutler, R. Konrat, E. Stupperich, G. Färber, K. Gruber and C. Kratky, Inorg. Chem., 1994, 33, 4128.
65. E. Stupperich, I. Steiner and M. Rühlemann, Anal. Biochem., 1986, 155, 365.
66. J. A. Horig, P. Renz and G. Heckmann, J. Biol. Chem., 1978, 253, 7410.
67. P. Renz, Methods Enzymol., 1971, 18C, 82.
68. K. L. Brown, X. Zou, J. Li and G. Chen, Inorg. Chem., 2001, 40, 5942.
69. C. Schöneich and K.-D. Asmus, Radiat. Environ. Biophys., 1990, 29, 263.
70. C. Schöneich, K.-D. Asmus and M. Bonifacic, J. Chem. Soc., Faraday Trans., 1995, 91, 1923.
71. W. Karmann, A. Granzow, G. Meissner and A. Henglein, Int. J. Radiat. Phys. Chem., 1969, 1, 395.
72. J. Redpath, Radiat. Res., 1973, 54, 364.
73. J. A. Franz, B. A. Bushaw and M. S. Alnajjar, J. Am. Chem. Soc., 1989, 111, 268.
74. P. Huston, J. H. Espenson and A. Bakac, Inorg. Chem., 1992, 31, 720.
75. L. E. H. Gerards and S. Balt, Recl. Trav. Chim. Pays-Bas, 1992, 111, 411.
76. L. E. H. Gerards and S. Balt, Recl. Trav. Chim. Pays-Bas, 1994, 113, 137.
77. B. P. Hay and R. G. Finke, Polyhedron, 1988, 7, 1469.
78. C. D. Garr and R. G. Finke, J. Am. Chem. Soc., 1992, 114, 10440.
79. C. D. Garr and R. G. Finke, Inorg. Chem., 1993, 32, 4414.
80. H. P. C. Hogenkamp, J. Biol. Chem., 1963, 238, 477.
81. K. N. V. Duong, A. Gaudemer, M. D. Johnson, R. Quillivic and J. Zyber, Tetrahedron Lett., 1975, 34, 2997.
82. K. L. Brown and S. Peck-Siler, Inorg. Chem., 1988, 27, 3548.
83. K. L. Brown and G.-Z. Wu, Inorg. Chem., 1994, 33, 4122.
84. K. L. Brown, J. M. Hakimi, D. M. Nuss, Y. D. Mantejano and D. W. Jacobsen, Inorg. Chem., 1984, 23, 1463.
85. P. A. Levene and E. T. Stiller, J. Biol. Chem., 1933, 102, 187.
86. R. S. Klein, H. Ohrui and J. J. Fox, J. Carbohydr., Nucleosides, Nucleotides, 1974, 1, 265.
87. T. Mukaiyama, Y. Hashimoto, Y. Hayashi and S.-I. Shoda, Chem. Lett., 1984, 557.
88. S. DeBarnardo and M. Weigele, J. Org. Chem., 1976, 41, 287.
89. R. J. Blau and J. H. Espenson, J. Am. Chem. Soc., 1985, 107, 3530.
90. J. D. Brodie, Proc. Natl. Acad. Sci. USA, 1969, 62, 461.
91. J. H. Grate and G. N. Schrauzer, J. Am. Chem. Soc., 1979, 101, 4601.
92. K. L. Brown and H. B. Brooks, Inorg. Chem., 1991, 30, 3420.
93. M. D. Waddington and R. G. Finke, J. Am. Chem. Soc., 1993, 115, 4629.
94. K. L. Brown, S. Cheng and H. M. Marques, Inorg. Chem., 1995, 34, 3038.
95. K. L. Brown, L. Salmon and J. A. Kirby, Organometallics, 1992, 11, 422.
96. M. P. Jensen and J. Halpern, J. Am. Chem. Soc., 1999, 121, 2181.
97. K. L. Brown, J. M. Hakimi and D. W. Jacobsen, J. Am. Chem. Soc., 1984, 106, 7894.
98. H. J. C. Yeh, K. L. Kirk and L. A. Cohen, J. Chem. Soc., Perkin Trans. 2, 1975, 928.
99. H. M. Marques, T. J. Egan, J. H. Marsh, J. R. Mellor and O. Q. Munro, Inorg. Chim. Acta, 1989, 166, 249.

---